18.07.2014 Views

The entirely mixing variables method

The entirely mixing variables method

The entirely mixing variables method

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

<strong>The</strong> <strong>entirely</strong> <strong>mixing</strong> <strong>variables</strong> <strong>method</strong><br />

Pham Kim Hung<br />

In some situations, proofs of conditional inequalities by making two <strong>variables</strong> equal<br />

become difficult, even impossible. It often happens with cyclic problems, when the<br />

equality holds for totally different <strong>variables</strong>. <strong>The</strong>re is a very helpful <strong>method</strong> which is<br />

often used to resolve this matter called the <strong>entirely</strong> <strong>mixing</strong> <strong>variables</strong> <strong>method</strong>. It is for<br />

annulling one variable that’s equal to 0, namely subtracting simultaneously from all<br />

<strong>variables</strong> a fixed value. <strong>The</strong> <strong>method</strong> can be used when there is a disparity of terms that<br />

approximate to 0 (which can be understand is three differences (a − b), (b − c), (c − a)<br />

for three-variable problems).<br />

See the following example to know thoroughly about it<br />

Problem 1. Prove that for all non-negative real numbers a, b, c we have<br />

a 3 + b 3 + c 3 − 3abc ≥ 4(a − b)(b − c)(c − a).<br />

Solution. <strong>The</strong> above inequality can be rewritten as follow<br />

(a + b + c) ( (a − b) 2 + (b − c) 2 + (c − a) 2) ≥ 8(a − b)(b − c)(c − a) (1)<br />

WLOG, we may assume that c = min(a, b, c). Fix the differences a − b, b − c, c − a<br />

and decrease simultaneously from a, b, c one value c (which means that we supersede<br />

a, b, c by a − c, b − c, 0), hence a − b, a − c, b − c don’t change but a + b + c is decreased.<br />

Thus the left hand side of (1) is decreased but the right hand side of (1) is invariable.<br />

So we only need to prove the problem in case a, b ≥ c = 0, the problem becomes<br />

a 3 + b 3 ≥ 4ab(b − a),<br />

But this above one is obviously true because<br />

4a(b − a) ≤ b 2 ⇒ 4ab(b − a) ≤ b 3 ≤ a 3 + b 3 .<br />

<strong>The</strong> equality occurs if and only if a = b = c. ❑<br />

This is only a simple sample, but represent the idea of <strong>method</strong> straightforwardly.<br />

A remarkable feature of this solution is a small comment, that if we decrease a, b, c<br />

at once by a number which is not bigger than min(a, b, c), one side is decreased but<br />

another side is immutable. To prove it is very easy, because the degree (of close-to-0<br />

values) of (a + b + c) ( (a − b) 2 + (b − c) 2 + (c − a) 2) is 2, and of (a − b)(b − c)(c − a)<br />

is 3. We also have a similar problem of four <strong>variables</strong> as follow<br />

Problem 2. Prove that for all non-negative real numbers a, b, c, d then<br />

a 4 + b 4 + c 4 + d 4 − 4abcd ≥ 2(a − b)(b − c)(c − d)(d − a),<br />

and in case (a − b)(c − d) ≤ 0, the inequality above is still true if replacing 2 by 17.<br />

Mathematical Reflections 5 (2006) 1


Solution. Firstly, we will prove the inequality in case (a − b)(c − d) ≤ 0 and since<br />

then, the reader can easily conclude all parts of the problem with the same manner.<br />

If d = 0 hence a ≤ b, the problem becomes<br />

a 4 + b 4 + c 4 ≥ 17ac(b − a)(b − c).<br />

If b ≤ c, we have done. Otherwise, if b ≥ a, c, by AM − GM Inequality<br />

ac(b − a)(b − c) ≤ ac(b − t) 2 ≤ t 2 (b − t) 2 ,<br />

for which t = (a + c)/2 ≤ b. Besides a 4 + c 4 ≥ 2t 4 , so it suffices to prove that<br />

2t 4 + b 4 ≥ 17t 2 (b − t) 2 .<br />

Let x = b/t − 1 ≥ 0, the above inequality becomes<br />

2 + (x + 1) 4 ≥ 17x 2<br />

⇔ x 4 + (x − 1)(4x 2 − 7x − 3) ≥ 0<br />

⇔ 1 ≥ y(1 − y)(3y 2 + 7y − 4).<br />

Clearly, if y ≥ 1 we have done. Otherwise, if y ≤ 1 we have two cases<br />

+, If 3y 2 + 7y ≤ 8 then RHS ≤ 4y(1 − y) ≤ 1 of course.<br />

+, If 3y 2 + 7y ≥ 8 then y ≥ 0.8, therefore<br />

y(1 − y) ≤ 0.8(1 − 0.8) ≤ 1 6 ⇒ y(1 − y)(3y2 + 7y − 4) ≤ 1 6 (3y2 + 7y − 4) ≤ 1.<br />

So the problem is proved if d = 0. Suppose that a, b, c, d > 0, d = min(a, b, c, d), then<br />

LHS = a 4 + b 4 + c 4 + d 4 − 4abcd = (a 2 − c 2 ) 2 + (b 2 − d 2 ) 2 + 2(ac − bd) 2<br />

= 1 2(<br />

(a − c) 2 (a + c) 2 + (b − d) 2 (b + d) 2 + (a − b) 2 (c + d) 2 +<br />

+ (c − d) 2 (a + b) 2 + (a − d) 2 (a + d) 2 + (b − c) 2 (b + c) 2) .<br />

Certainly, this step claims that d = 0 is all work to complete, it ends the proof. ❑<br />

In some instances, when we decrease simultaneously all <strong>variables</strong>, both hand sides<br />

of the inequality are changed simultaneously too, increasing or decreasing. Consider<br />

the following examples which are more difficult<br />

Problem 3. Let a, b, c be non-negative real numbers with sum 3. Find all possible<br />

values of k for which the below inequality is always true<br />

a 4 + b 4 + c 4 − 3abc ≥ k(a − b)(b − c)(c − a).<br />

2


Solution. In case c = 0, easy to prove that −6 √ 2 ≤ k ≤ 6 √ 2. Indeed, if c =<br />

0, a + b = 3, applying AM − GM Inequality, we obtain<br />

LHS = a 4 + b 4 = (a 2 − b 2 ) 2 + 2a 2 b 2 ≥ 2 √ 2|ab(b 2 − a 2 )| = 6 √ 2|ab(b − a)|.<br />

<strong>The</strong> equality is taken if and only if |b 2 − a 2 | = |ab| and a + b = 3, c = 0.<br />

Because the equality can be happened if a, b, c are different values, so the positive<br />

value of k to contruct a valid inequality must lie between −6 √ 2 and 6 √ 2.<br />

Now we will prove that for all non-negative real numbers a, b, c adding up to 3 then<br />

a 4 + b 4 + c 4 − 3abc ≥ 6 √ 2(a − b)(b − c)(c − a).<br />

To perform the idea of the <strong>entirely</strong> <strong>mixing</strong> <strong>variables</strong> <strong>method</strong>, the first neccesary thing<br />

is normalizing two sides of the inequality<br />

a 4 + b 4 + c 4 − abc(a + b + c) ≥ 2 √ 2(a − b)(b − c)(c − a)(a + b + c).<br />

Decreasing or increasing merely all <strong>variables</strong> by a number t ≤ min(a, b, c), we only<br />

realize that both sides of the above inequality are changed. But the feature here is<br />

that we can compare these changing values. Indeed, consider the function<br />

f(t) =<br />

(a + t) 4 + (b + t) 4 + (c + t) 4 −<br />

− (a + t)(b + t)(c + t)(a + b + c + 3t)<br />

− k(a − b)(b − c)(c − a)(a + b + c + 3t)<br />

⇒ f(t) = A + Bt + Ct 2 ,<br />

in which the coefficients of f(t) are<br />

A = a 4 + b 4 + c 4 − abc(a + b + c) − k(a − b)(b − c)(c − a),<br />

B = 4(a 3 + b 3 + c 3 ) − (a + b + c)(ab + bc + ca) − 3abc − k(a − b)(b − c)(c − a),<br />

C = 6(a 2 + b 2 + c 2 ) − (a + b + c) 2 − 3(ab + bc + ca).<br />

Clearly, C ≥ 0 because<br />

6(a 2 + b 2 + c 2 ) ≥ 2(a + b + c) 2 ≥ (a + b + c) 2 + 3abc.<br />

WLOG, assume that c = min(a, b, c). We will prove<br />

4(a 3 + b 3 + c 3 ) − (a + b + c)(ab + bc + ca) − 3abc ≥ 6 √ 2(a − b)(b − c)(c − a).<br />

<strong>The</strong> left hand side can be analyzied to sum of squares as follow<br />

LHS = (a − b) 2 (a + b) + (b − c) 2 (b + c) + (c − a) 2 (c + a) + 2(a 3 + b 3 + c 3 − 3abc)<br />

= (a − b) 2 (2a + 2b + c) + (b − c) 2 (2b + 2c + a) + (c − a) 2 (2c + 2a + b).<br />

Mathematical Reflections 5 (2006) 3


From this transformation, we only need to examine one case c = 0 (that is based on<br />

the first idea of the <strong>entirely</strong> <strong>mixing</strong> <strong>variables</strong> <strong>method</strong>). If c = 0, we need to prove that<br />

4(a 3 + b 3 ) − ab(a + b) ≥ 6 √ 2ab(b − a)<br />

⇔ a 3 + 4b 3 + (6 √ 2 − 1)a 2 b ≥ (6 √ 2 + 1)ab 2 .<br />

But this last one is obviously true by AM − GM Inequality<br />

4b 3 + (6 √ √<br />

2 − 1)a 2 b ≥ 4 6 √ 2 − 1ab 2 ≥ (6 √ 2 + 1)ab 2 .<br />

Thus in the expression of f(t), the coefficients of t, t 2 are positive at once, it implies<br />

that f(t) is an increasing function for t ≥ 0. If c = min(a, b, c), we have<br />

a 4 + b 4 + c 4 − abc(a + b + c) − k(a − b)(b − c)(c − a)(a + b + c)<br />

≥ a ′ 4 + b<br />

′4 + c<br />

′4 − a ′ b ′ c ′ (a ′ + b ′ + c ′ ) − k(a ′ − b ′ )(b ′ − c ′ )(c ′ − a ′ ),<br />

in which a ′ = a − c, b ′ = b − c, c ′ = c − c = 0. It implies that we only need to check the<br />

first problem in case c = 0, which was showed as above. <strong>The</strong> problem is completely<br />

solved and the equality holds for a = b = c or the following case with its permutations<br />

c = 0, b = 1 + √ 5<br />

2<br />

a ⇔ a = 3(3 − √ 5)<br />

2<br />

, b = 3(√ 5 − 1)<br />

, c = 0. ❑<br />

2<br />

This is clearly a painstaking and complicated solution, but the main content is<br />

only involved by the idea of the <strong>entirely</strong> <strong>mixing</strong> <strong>variables</strong> <strong>method</strong>. Using functions as<br />

above is almost a helpful way for problems of this kind. One special example is Jack<br />

Grafukel’s Inequality, where this <strong>method</strong> shows the most original, simplest solution<br />

Problem 4. Prove that for all non-negative real numbers a, b, c then the following<br />

inequality holds<br />

a b c<br />

√ + √ + √ ≤ 5 a + b + c.<br />

a + b b + c c + a 4√<br />

(Jack Grafulkel, Crux)<br />

Solution. Firstly, we must prove the problem in case one of three numbers a, b, c is<br />

0. Indeed, suppose c = 0, the given inequality can be changed into<br />

a<br />

√<br />

a + b<br />

+ √ b ≤ 5 4√<br />

a + b<br />

⇔ 1 4 a + 5 4 b ≥ √ b(a + b)<br />

⇔ (a + b) + 4b ≥ 4 √ b(a + b).<br />

By AM − GM Inequality, it’s obviously true and has equality when a = 3b.<br />

4


Next, we will solve this problem in the general case. Denote<br />

√ √ √<br />

a + b a + c b + c<br />

x =<br />

2 , y = 2 , z = 2 , k = 5√ 2<br />

4 .<br />

WLOG, we may assume that x = max(x, y, z). <strong>The</strong> required inequality is equivalent<br />

to<br />

y 2 + z 2 − x 2<br />

+ z2 + x 2 − y 2<br />

+ x2 + y 2 − z 2<br />

≤ 5√ 2√ x + y + z<br />

z<br />

x<br />

y 4<br />

( 1<br />

⇔ x + y + z + (x − y)(x − z)(z − y)<br />

xy + 1<br />

yz + 1 )<br />

≤ k √ x<br />

zx<br />

2 + y 2 + z 2 (1)<br />

Clearly, the problem is proved in case z ≥ y. Firstly, we will prove that, for every<br />

numbers t ≥ 0 then<br />

k √ (x + t) 2 + (b + t) 2 + (z + t) 2 ≥ k √ x 2 + y 2 + z 2 + 3t (2)<br />

⇔ k √ x 2 + y 2 + z 2 + 2t(x + y + z) + 3t 2 ≥ k √ x 2 + y 2 + z 2 + 3t.<br />

Note that k 2 ≥ 3 and x + y + z ≥ √ x 2 + y 2 + z 2 so the above inequality is true.<br />

On the other hand, since x = max(x, y, z) and x 2 ≤ y 2 + z 2 , there exists a positive<br />

number t ≤ min(a, b, c) for which (x − t) 2 = (y − t) 2 + (z − t) 2 . Hence, as the above<br />

solved result, (there is one of three numbers a, b, c is 0) inequality (1) is true if we<br />

replace x, y, z by x ′ = x − t, y ′ = y − t, z ′ = z − t. So we obtain<br />

( 1<br />

x ′ + y ′ + z ′ + (x ′ − y ′ )(x ′ − z ′ )(z ′ − y ′ )<br />

x ′ y ′ + 1<br />

y ′ z ′ + 1 )<br />

z ′ x ′ ≤ k √ x ′2 + y ′2 + z ′2 (3)<br />

<strong>The</strong> inequality (1) can be rewritten as follow<br />

(<br />

(x ′ − y ′ )(x ′ − z ′ )(z ′ − y ′ )<br />

1<br />

(x ′ + t)(y ′ + t) + 1<br />

(y ′ + t)(z ′ + t) + 1<br />

(z ′ + t)(x ′ + t)<br />

+x ′ + y ′ + z ′ + 3t ≤ k √ (x ′ + t) 2 + (y ′ + t) 2 + (z ′ + t) 2 ,<br />

)<br />

+<br />

but clearly, we obtain this one directly by adding (3) and (2) (in which x, y, z were<br />

replaced by x ′ , y ′ , z ′ ). <strong>The</strong> inequality is completely solved and has equality when<br />

x = 3t, y = t, z = 0 or permutations. ❑<br />

Moreover, here is an similar example<br />

Problem 5. Let a, b, c be three side-lengths of a triangle. Prove that<br />

( ) a<br />

2<br />

2<br />

b + b2 c + c2 ≥ a + b + c + b2<br />

a<br />

a + c2 b + a2<br />

c .<br />

Solution. Clearly, this one is equivalent to<br />

∑ (a − b) 2<br />

≥<br />

b<br />

cyc<br />

(a − b)(b − c)(c − a)(a + b + c)<br />

.<br />

abc<br />

Mathematical Reflections 5 (2006) 5


⇔ ∑ ac(a − b) 2 ≥ (a − b)(a − c)(b − c)(a + b + c).<br />

cyc<br />

<strong>The</strong> above form shows that we only need to prove it in case a ≥ b ≥ c and a = b + c<br />

(indeed, we only need to prove ∑ cyc(a + c)(a − b) 2 ≥ 3(a − b)(a − c)(b − c), applying the<br />

<strong>mixing</strong> <strong>variables</strong> <strong>method</strong> again, it remains to prove that a(a − b) 2 + b 2 (b + a) + a 2 b ≥<br />

3ab(a − b), which is obvious). So we only need to prove the initial problem in case<br />

(a, b, c) are three lengths of a trivial triangle when a = b + c. <strong>The</strong> inequality becomes<br />

2 ( (b + c) 3 c + c 3 b + b 2 (a + b) ) ≥ 2bc(b + c) 2 + (b + c) 3 b + b 3 c + c 2 (b + c)<br />

⇔ b 4 − 2b 3 c − b 2 c 2 + 4bc 3 + c 4 ≥ 0.<br />

Because of the homogeneity, we may assume c = 1 and prove f(b) ≥ 0 for<br />

f(b) = b 4 − 2b 3 − b 2 + 4b + 1<br />

By derivative, it’s easy to prove this property. This ends the proof. ❑<br />

For the end, the reader should try proving two hard and interesting inequalities<br />

Problem 6. Let a, b, c be three side-lengths of a triangle. Prove that<br />

( a<br />

b + b c + c ) ( b<br />

a − 3 ≥ k<br />

a + c b + a )<br />

c − 3<br />

where k = 1 −<br />

2<br />

(<br />

2<br />

√<br />

2 − 1<br />

) √<br />

5 + 4<br />

√<br />

2 + 1<br />

.<br />

Problem 7. Prove that for all non-negative real numbers a, b, c with sum 1 then<br />

1 ≤<br />

a<br />

√<br />

a + 2b<br />

+<br />

b<br />

√<br />

b + 2c<br />

+<br />

c<br />

√ c + 2a<br />

≤<br />

√<br />

3<br />

2 .<br />

<strong>The</strong> detailed proof will be saved up for the reader. Try it !<br />

6

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!